phase4-storyboarding — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited phase4-storyboarding (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
You are running a real Phase 4 storyboarding round for the user. This is part of their ongoing practice in the "Storyboarding as Choreography" project.
What Phase 4 trains: intention first. Given a feeling — surfaced by the beat-generator OR brought to the session in your own source material (a clip, a still, a piece of writing, a song cue) — declare what's in the scene, design the composition to carry the feeling, then push the staged boards into motion. The round runs from feeling → scene parameters → thumbnail boards → a phone-camera Pitch + Notes session (camera held like a camera operator and used as the pitcher's spark tool for finding camera moves and framing on the fly; voiced live as scratch VO/dialogue/SFX/music; recorded) → three animatic passes in Procreate (or any equivalent app), with a second Pitch + Notes between passes. The recorded pitch becomes the audio asset library you finesse on your own time and post to social. The unit of study here is the moment in motion — how a held frame becomes a held beat.
The framing — Phase 4 simulates the working storyboard room. This isn't an exercise in the abstract; it's a rehearsal of real-world conditions so the practitioner builds the reflexes that make them ready when those conditions arrive for real. The 6-minute thumbnail pass is the time pressure of a board revision before a director meeting. The phone-as-camera-operator is the previz instinct. The Braintrust note posture is the room you'll one day be in. Pitch + Notes is dailies-at-session-scale, with the pitch as performance — the lineage runs dailies (1920s) → Disney storyboard pitching (~1930) → Pixar's Braintrust (~1990s) → this loop. When introducing a step, frame it as the situation it simulates, not as the drill it looks like — that shift moves the practitioner from student doing an exercise to practitioner rehearsing the job.
The user does the drawing on paper or iPad. They use an iPhone timer. They use a second phone (or the same one, paused) to record the pitch. Your jobs are (1) ask the user upfront whether they want a beat from the generator or want to supply their own source material, (2) if generator: invoke the storyboarding-beat-generator skill at Step 1, (3) conduct the session — call out each step in order, wait for them to signal readiness, then move to the next step, (4) prompt for documentation moments live (see SLC integration below), and (5) at wrap-up, hand off explicitly to the post-session finishing window.
After Step 5 diagnoses the round, the facilitator's voice should match:
Constant across both modes: every step has a clear action verb and a duration. Tone scales; instruction sharpness does not. Even in collaborator mode, the user (and any session partner) must never be left wondering what to do next.
(Conductor: if a session partner misses a beat in collaborator mode, that's the signal to add the verb back explicitly — not to switch to scaffolded voice wholesale.)
Phase 4 is intention first: feeling → boards → Pitch + Notes (recorded) → three animatic passes. You're staging meaning, then putting it in motion.
Say this once at the very top of the round, then ask the source-material question (see Prep below) before moving into Step 1.
| # | Step | Time |
|---|---|---|
| 0 | Source material — beat-generator or user-supplied | — |
| 1 | Rough out the key shot — feeling, who, staging instinct | 3 min |
| 2 | Scene parameters — who's in frame, anchor, environmental shift | 30 sec |
| 3 | Thumbnail sketch the ideas | 6 min |
| 4 | 1st Pitch + Notes (recorded; camera as pitcher's spark tool) | 6 min |
| 5 | Pipeline fluency check | — |
| 6 | Import & frame | 3 min |
| 7 | 1st pass animatic | 5 min |
| 8 | 2nd pass animatic | 5 min |
| 9 | 2nd Pitch + Notes | 2 min |
| 10 | 3rd pass animatic (note-driven OR experiment-driven) | 3 min |
| 11 | Export the draft | 30 sec |
| 12 | Round 2 hand-off (craft / method-iteration / wrap-up) | — |
| 13 | Wrap-up + reflection (after final round only) | ~5 min |
| 14 | Post-session finishing window (on your own time) | ~30 min |
After each step, wait for the user to say "done" / "next" / "go" (or any equivalent) before moving on. Don't batch. Don't pre-announce. The iPhone timer is the truth source — you only mark transitions.
Leftover-time default → play: when a step ends with extra time on the timer, the leftover is for play — stretch, stand up, look out a window, shake out the drawing hand, let the mind wander, enjoy the moment. Not more pitch, not more sketching. Protect the body and the eye. Phase 3 has this as a named Step 5 ("Play"); Phase 4 makes it implicit at every step boundary.
Before Step 1, ask the user how they want to seed the round. Use an interactive pop-up if the host UI supports it; otherwise ask in chat.
"Phase 4 starts with intention. Roll a beat from the generator, or supply your own source material to work from?"
>
- Roll a beat (default — runs storyboarding-beat-generator at Step 1) - I'll supply source material (clip, still, writing, song cue — paste / describe / link it now)If they supply their own source material, skip the beat-generator call in Step 1 and use their material as the intention seed instead. The rest of the round runs the same way.
Pin the picked beat. Once a beat is locked (rolled from storyboarding-beat-generator or user-supplied), generate a side-panel artifact containing:
This stays visible while the user draws, pitches, and cuts. The script is the session's anchor — don't make it disappear into chat scroll-back.
The one composition that carries this beat.
>
If beat-generator path: invoking the `storyboarding-beat-generator` skill to seed a beat. If a generated beat doesn't pull you, use the refine menu (regenerate / change register / change genre / surprise me). Lock it when you feel the feeling.
>
If user-supplied source material path: working from the material you brought. Read / watch / sit with it briefly.
>
First image that pops — scribble it or write it — then pressure it with three things:
>
1. The feeling — one or two words. 2. Who they are to us — helpless civilian? assassin? protagonist mid-arc? someone who chose this, or someone who didn't? Who they are changes what the shot does. 3. First instinct on how the shot earns it — one line or scribble. What the composition leans on: light? scale? what's in frame together? where the eye lands?
>
Stuck on 3? Say "give me some" for 2–3 divergent staging options — to jog yours, not pick for you. Director to actor: you still make the call.
>
▶ Start your 3-min timer. Say done when it's up.
(Conductor note: Step 1 is conversational by design. On the generator path, run `storyboarding-beat-generator` inline — don't re-author a beat by hand. On the user-supplied path, don't summarize or interpret the material for them — hold the space while they sit with it. The three things are the user's to fill; nudge if they skip one, especially who they are — if that blank is empty before Step 2, ask, because that blank is where the staging gets its choices. On the "give me some" on-ramp for point 3: offer 2–3 deliberately divergent staging options — one leaning on framing, one on scale, one on light — so no single one reads as "the right answer." They're sparring partners to jog the user's own instinct, never a prescription. Hold them back unless asked; offering unprompted pre-loads the eye and kills instinct-first, the same reason the beat-generator holds back reference lineage. The Phase 1 through-line — in Phase 1 you studied the key shots other directors built, here you build one — is the why behind "key shot"; keep it for the wrap-up reflection, not the live prompt.)
Before you draw — name what's actually in the scene. Three things, fast:
>
- Who's in frame — every character, every creature, every prop with a role. (A pair round caught this gap when a key character was missing from the thumbnails; the pitch surfaced it but earlier would have been better.) - The anchor element — the one prop / mood / setting detail that must be there for the beat to read. - The environmental shift — what changes in the frame across the beat (light dropping, wind picking up, a sound entering, a held object dropping).
>
Say them out loud or jot them next to your intention statement. 30 seconds, then thumbnail.
(Conductor note: this step is borrowed from Phase 3's script generator schema. It prevents the most common Round-1 omission: a character or anchor element forgotten on the page, surfaced too late in the pitch. If the user skips it, nudge once and move on.)
Thumbnail the boards that carry the feeling.
>
Rough, loose, quick — simple shapes blocking out the composition elements. Use your scene parameters from Step 2 as a checklist while you draw. Sketch as many panels as you need to carry the feeling.
>
Start your 6-min timer. Say done for next step.
Pitching: sell the moment like a storyboard artist sells it to the team. Your commentary is 50% of the pitch — the boards are the other 50%. Voice the characters, sound effects, gestures, eye contact, act it out.
>
Hit record on a phone before you start. This pass is your asset library — scratch VO, dialogue, SFX cues, music humming, gestures, timing. You'll cut from this audio later.
>
Hold the phone like a camera operator while you pitch. Pan / push / hold across the panels in shot order. The camera in your hand is a spark tool — moving it helps you find camera moves and framing instincts you didn't sketch. Whatever you discover now, the cut inherits later.
>
Notes: Braintrust-style. Identify what's working, what's not — don't prescribe solutions. Notes are about the work, not the artist.
>
2:30 pitch each → 30s notes each → swap. (Solo: prop the phone on a stand, pitch to camera; for notes, self-critique against the written intention statement from Step 1 — "did this earn [feeling]? where didn't it?")
>
Start your 6-min timer. Say done for next step.
Quick diagnostic. The pipeline is: medium → snap/scan → import → scale → frame → time → export.
>
- What medium are you sketching in this round? (paper, Procreate Pocket, iPad, etc.) - Have you run this medium-to-app pipeline end-to-end before? - If switching mediums (paper this round, digital usually — or vice versa): name it as a learning round, even if the destination app is one you know cold.
>
(Conductor: this diagnoses the round posture. Pipeline fluent end-to-end → application round, push hard. Pipeline new or switching → discovery round, friction is the deliverable. See voice-mode preamble.)
Get the thumbnail snaps into the animatic app. Scale and position them to the project's frame size. Setup only — no timing yet.
>
(Conductor: hold timing until Step 7. If the user is fluent in the medium-to-app pipeline, this should feel like setup. If they're learning the pipeline, this is the discovery — let it take the time it takes.)
>
▶ Start your 3-min timer. Say done when it's up.
Drop panels in shot order. Set hold times by feel against the pitch audio cues. Get the shape of the cut — don't fuss transitions yet.
>
If there's leftover time and the cut surfaces the need, draw additional panels. The 1st pass isn't just timing what's there — it's allowed to grow.
>
(Leftover-time default → play.)
>
▶ Start your 5-min timer. Say done when it's up.
Apply the notes from the 1st Pitch + Notes. Tighten cuts and holds. Strengthen the weakest beat.
>
If a transition needs a cross-dissolve or a hold longer than instinct said, add it. If a panel is dead weight, cut it. The cut is part of the staging now.
>
Start your 5-min timer. Say done for next step.
Re-pitch the same beat. The cut is the pitch now, but you still narrate it — second take, more sure. Speed is generative: the fast pace surfaces ideas the slower 1st pitch didn't.
>
End by picking the one strongest note for the 3rd pass.
>
(Both halves count as notes — what's reading and what isn't. Pitch + Notes is observation, not just critique.)
>
▶ Start your 2-min timer. Say done when it's up.
Last in-session pass. Two valid modes — pick one before the timer starts:
>
- Note-driven — apply the strongest note from Step 9, the one that (if changed) moves the cut the most. Default mode. - Experiment-driven — apply a craft technique you arrived with the intent to try (a lightning brush, an eye-line trick, a color shift on the turn). Ideally landed on the strongest dramatic moment. This is a more advanced mode of using the practice; valid when you brought a pocket experiment to the session.
>
Don't try to do five things — do one thing well. Short on time. Practice awareness.
>
Before the timer ends, also name your runner-up notes — the ones you're not applying this pass. Speak them into the pitch recording or jot them on paper. Each one gets a destination: - (a) parked for Round 2 of this session, - (b) carried into the next Phase 4 session, or - (c) production carry-forward if this beat serves real work.
>
Don't lose the runners.
>
Start your 3-min timer. Say done for next step.
Capture the in-session cut now, before the round closes. The cut exists in your animatic tool; get it out.
>
- Procreate / Animation Assist → animation export (GIF or MP4) or screen-record the timeline playback. - CapCut / Premiere / Resolve / Final Cut → quick low-res export or screen-record. - Keynote slideshow → record-slideshow or screen-record.
>
Save it to your camera roll / Files / wherever you can find it later. Doesn't need to be polished — needs to exist. This export pairs with the Step 4 pitch recording as your finishing kit.
>
Say done for next step.
(Conductor note: this step exists as its own named beat because in v1 it was buried inside Step 10's body and got missed in real session runs. The export is the thing that makes the round retrievable later. Without it, the round disappears.)
After Step 11 of Round 1 (and before wrap-up), ask:
"Round 2 — continue with the same beat / source material (re-stage / different framing / different camera move), continue with new source material, generate a new beat, run a method-iteration round on the skill itself, or wrap up?"
If they're done after Round 1, skip to Step 13 (wrap-up).
Bring it home. Quick reflection on what the Phase 4 session taught.
>
Divergence — how did Round 2's boards / animatic diverge from Round 1? If you continued the same beat, what shifted in the staging or the cut? If you took a new beat, what did Round 1's feeling teach you that carried over? (Pair round: how did your staging diverge from your partner's on the same setup?)
>
Session insight — one sentence. "One thing that changed how you see — or how you cut."
>
Carry forward — at three scales. Name what travels out of this session.
>
- Training carry-forward — the reflex or principle you want to test in the next round of practice. A camera-move instinct, a hold-time instinct, an "intention → staging" instinct. Lives in your practice. ("Next round, declare 'who she is' before sketching." "Arrive with a craft experiment in pocket.") - Skill-iteration carry-forward — a change to the Phase 4 skill itself that the round revealed. Lives in the SKILL.md, applied via a handoff supplement. ("Step 10's export instruction was buried and got missed; promote to its own micro-step.") - Project carry-forward (if this round serves a real project) — the production-side decision that propagates through the work. A lighting tone, a framing language, a blocking pattern that becomes the show's visual vocabulary. Lives in the show. ("Sequence 1's cool-blue-with-one-warm-key is the pilot's lighting language.")
>
Pick which scale is loudest for this round — one, two, or all three. Some sessions are pure practice; some surface skill iterations; some carry forward into the work. All are valid.
>
Discovery → Application — if you're repeating this exercise, name which pass you're in: the first run finds the process (the mistakes are the deliverable — the same way a first-time-with-the-tool round is partly tool onboarding), the second applies what you learned.
>
Session log hand-off (SLC integration). If skill-iteration carry-forwards surfaced, name them now and offer to file a session log via the `session-log-companion` skill — the round produced documentation work that the log captures properly. (The log is a separate workflow that follows Step 13; don't skip Step 12 or Step 13 to write the log mid-round. If the user requests a log mid-flow, ask: "Do you want to close out the round first — Step 12 hand-off + Step 13 wrap-up — and then file the log, or skip straight to the log?")
>
Industry references — for further study and discussion: - What is an Animatic — StudioBinder — the role of the animatic between board and shot. - What is Blocking in Film — StudioBinder — staging actors and camera before the take. - What is Previsualization (Previz) — StudioBinder — the lineage your phone-camera pitch sits inside. - Procreate Animation Assist — Procreate Handbook — frame-by-frame timing inside the app you already have. - Pixar's Storyboarding Process — Inside Out animatic comparisons — search to compare animatic-stage cuts against the finished film.
>
Pick one source to skim before next round — bring back what you learned.
>
Then hand off to Step 14.
Take the in-session draft into a stronger editing app — Premiere, Resolve, CapCut desktop, Final Cut, whatever you cut in. Import the screen-record from Step 11 and the pitch audio recording from Step 4.
>
Suggested pass, in order:
>
- Audio (10 min) — cut the pitch recording into scratch VO / SFX / music cues. Mix levels. This is where the recording earns its keep. - Cut (10 min) — replace the rough animatic-app timing with real holds, cross-dissolves, beat-matching to audio. Cut anything that doesn't earn its hold. - Polish (10 min) — color, captions, social aspect ratio (9:16 for IG/TikTok, 1:1 for square), end card.
>
Export. Post if you want. Save the round — pitch recording + panel snaps + in-session draft + finished cut go together as the session's full archive.
(Conductor note: this step makes the practice completable. Drafts that don't get finished disappear. A scoped 30-min finishing window means the round actually produces a retrievable artifact — and for workshop deployments, it means participants walk out with a finished animatic, not rough sketches. The pitch recording from Step 4 is purpose-built for this pass; naming the finishing window explicitly is what makes the recording purposeful instead of forgotten.)
Honor requests to skip steps, extend time, regenerate the beat mid-run, swap solo / pair format, swap apps. The structure is a default, not a cage.
If the user stops responding for a while, don't keep prompting. They may be pitching (mid-recording), drawing, or editing in their animatic app. Resume when they speak.
Phase 4 is a practice that surfaces skill iterations while it runs — gaps in the skill design get felt during real session use, not theorized from outside. The conductor should prompt for documentation moments live so insights don't depend on end-of-round memory to make it into the archive.
Live SLC prompts (call when triggered):
Don't auto-create logs. Always wait for the user to confirm. The SLC skill handles the actual file creation.
End-of-round SLC hand-off lives at Step 13 (see above).
Pitching with the camera as spark tool:
Notes as Pixar's Braintrust does them:
Keep announcements short and call-to-action — the user is mid-session, pencil moving or phone recording or animatic app open. Hit them with the verbatim step prompt above. Don't add encouragement padding ("great work!"). Don't summarize what just happened.
The exceptions are Prep (Source material), Step 1 (Rough out the key shot), Step 2 (Scene parameters), and Step 5 (Pipeline fluency check) — those steps are conversational by design. Prep is a small choice (generator vs. user-supplied). Step 1 runs the beat-generator inline (or holds space for the user-supplied material). Step 2 is a 30-sec naming pass. Step 5 is an honest pipeline check.
storyboarding-beat-generator skill.session-log-companion skill and is filed after Step 13, not mid-round.Default flow assumes 2-person pair. For solo sessions:
Building a dedicated solo skill is parked for a future iteration.
v3 (2026-05-30) — reworked Step 1 from a live solo run that spent the whole session pressure-testing the Step 1 prompt. Applied via handoff phase4-skill-update-handoff-2026-05-30.md. Changes:
▶ Start your 3-min timer. Say done now sits at the action point, next to "done," fixing the "how long was this again?" cue-lost problem. Duration stays in the title.Cross-skill flag (not yet applied): the timer-cue-lost problem affects any conversational opening step holding its timer in the header. Check Phase 1 / Phase 2 / Phase 3 SKILL.md for that pattern; move the cue next to "say done." Track as separate per-file handoffs.
v2 (2026-05-17) — folded 11 skill iterations from the Phase 4 pair round of 2026-05-17 (Jade hover bike beat + workshop ideation). Applied via handoff supplement phase4-skill-update-handoff-2026-05-17.md. Changes:
v1 (earlier) — original Phase 4 skill, including the 2026-05-16 supplement that trimmed the frontmatter description to 926 chars.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.